Human Preference Dataset

该数据集是从Stable Foundation Discord频道收集的人造图像上的人类选择，旨在理解人类偏好并提升文本到图像生成的技术水平。该数据集包含了2,659位不同用户所做的选择，涵盖了多种主题的高度多样性。同时，所有包含不适宜内容（NSFW）的图像已被筛选掉。该数据集规模宏大，包含98,807张由25,205个提示生成的图像，其任务是对文本到图像生成技术进行评估。

This dataset comprises human selections on synthetic images collected from the Stable Foundation Discord channel, with the objective of understanding human preferences and advancing the state of the art in text-to-image generation technologies. It includes selections made by 2,659 distinct users, covering high diversity across a wide range of topics. Meanwhile, all images containing inappropriate (NSFW) content have been filtered out. This large-scale dataset contains 98,807 images generated from 25,205 prompts, and its purpose is to evaluate text-to-image generation technologies.